Comparison of rates of dementia among older adult recipients of two, one, or no vaccinations

INTRODUCTION

Animal and human studies have provided support for infectious etiologies of Alzheimer's disease (AD) and dementia.1 Viral infection may exacerbate inflammation and oxidation in the aging brain and precipitate cerebrovascular damage leading to dementia.2-8 Treatment and/or prevention of viral infection may reduce neurocognitive impairment through this mechanism.4, 5 With respect to prevention, a range of vaccines, including influenza, herpes zoster (HZ) and tetanus, diphtheria and pertussis (Tdap), are associated with a reduced dementia risk.9-13

Training of the immune system with a lifetime of vaccinations may lower dementia risk by decreasing neuroinflammation. This may explain the dose–response relationship observed in an analysis of Taiwanese National Health data. Compared with unvaccinated patients, the risk of dementia decreased with increasing number of annual influenza vaccinations.13 Our own analysis of Veterans Health Administration (VHA) data suggests reduced dementia risk requires ≥6 annual influenza vaccines as compared with none.14 However, accurate measurement of influenza vaccination status in the United States is a challenge because many persons obtain vaccinations outside their healthcare system,15 and we lack a national immunization registry.16 Overall, the dose–response relationship between vaccination and dementia has not been well described.

The present study was designed to first determine if there is a dose–response relationship between the number of different adult vaccinations received and incident dementia among patients ≥65 years of age. We determined if patients who had both the HZ and Tdap vaccinations had a lower risk for dementia than those with HZ alone, Tdap alone, or neither. Second, we employed an active comparator design to test whether dementia risk differed between patients who received only HZ versus only Tdap. Third, we conducted analysis in a VHA patient cohort with replication in a private sector MarketScan claims database.

METHODS Design

This was a secondary analysis of de-identified data and was deemed non-human subjects research by the Saint Louis University Institutional Review Board. This retrospective cohort study followed STROBE reporting guidelines (Table S1).

Data sources

National VHA administrative medical record data included observation time from fiscal years (FY) FY09–FY19 (October 1, 2008–September 30, 2019). VHA data included ICD-9 and ICD-10 diagnostic codes, medication dispensing, current procedural terminology (CPT) codes, vital signs, laboratory results, type of clinic visit (e.g., primary care,), and demographic measures. VHA maintains Medicare claims and Part-D pharmacy claims linked to administrative medical record data. Medicare claims and Part-D pharmacy claims were used to obtain diagnoses codes, laboratory values, and prescription fills.

IBM® MarketScan® Commercial Claims and Medicare Supplemental databases covered January 1, 2009–December 12, 2018 (calendar year [CY]2009–CY2018) and contained the same measures available in VHA data, unless otherwise stated. The IBM® MarketScan® database included healthcare claims from ambulatory and hospital care, academic and nonacademic health systems, and patients had either private health insurance or Medicaid/Medicare.

Eligibility

Receipt of vaccinations and use of preventive health care are correlated17 and may contribute to dementia. To control for this bias, we first limited the VHA and MarketScan samples to patients with at least three well-visits in the observation period. As reported below, we also controlled for the number of well-visits in the 2 years prior to index.

In the VHA, index date or baseline was October 1, 2011 (FY2012), by which time patients must have been ≥65 years of age. In the 2 years prior to the index date, we excluded patients who did not have VHA encounters. Patients with prevalent dementia or dementia-related medications (e.g., donepezil, rivastigmine, galantamine, memantine) or those with conditions associated with memory impairment such as Creutzfeldt–Jakob disease in the 2 years prior to index were excluded. Minimum follow-up time was 91 days after index to allow for a biologically plausible association between vaccination and incident dementia. Dementia cases or non-cases censored in the first 90 days of follow-up were excluded. There were 195,375 eligible VHA patients at index date. Sample selection was similar in the MarketScan cohort, which resulted in 193,259 eligible patients at index date (January 1, 2012; CY2012).

We employed a per-protocol approach; therefore, patients' Tdap and HZ vaccination status was measured as receiving or not receiving a vaccine by index date, and those without a vaccination at index were required to remain without vaccination during follow-up. Also, the Tdap only group had to remain free of HZ vaccination in follow-up and vice versa. After removing patients with missing demographic measures, the VHA analytic sample contained 80,070 patients, and the MarketScan sample contained 129,200 patients. The sampling process is shown in Figure S1A,B.

Variable definitions

Detailed definitions of all variables are shown in Table S2.

Exposure

For persons ≥65 years of age, Tdap vaccination was defined by CPT codes 90701 or 90715 and by product names Adacel and Boostrix, and HZ vaccination was defined by CPT codes: 90710, 90716, 90736, 90750, and product names: Proquad, Varivax, Zostavax, and Shingrix. Vaccinations were measured using the CPT codes or pharmacy product name. Vaccination exposure was classified into (1) no vaccination, (2) Tdap vaccination only, (3) HZ vaccination only, and (4) Tdap and HZ vaccination. This approach to defining vaccinations has been shown to have excellent agreement with manual chart abstraction.18

Outcome

Incident dementia in follow-up was defined by ICD-9 and or ICD-10 diagnostic codes on two separate days in any 12-month period. The first of the two codes were used to date the diagnosis. This definition has good agreement with Mini Mental State Exam and the Saint Louis University Mental Status Examination scores indicating mild or worse dementia.19 We have applied this method to define dementia in prior studies of metformin and incident dementia20-22 and in studies of Tdap vaccination,11 influenza vaccination,14 and incident dementia.

Follow-up time

Follow-up time was defined as months from index date to dementia or censoring. Censoring in the VHA occurred at last available VHA encounter or Medicare claim, or death. Mortality data were not available in MarketScan, therefore censoring in MarketScan occurred at last available inpatient or outpatient claim.

Covariates

To compute propensity scores, covariates were measured in the 2 years prior to index date. Covariates included sociodemographics (age, gender, race-VHA only, and marital status-VHA only), insurance status (VHA only vs. VHA plus other source of insurance), geographic region, overall health care utilization, and use of well-visits in the 2 years prior to index. Thus, we both sampled on 3 or more well-visits and controlled for the number of well-visits 2 years before index. Comorbid conditions included cancer, type 2 diabetes, obesity, hypertension, stroke, ischemic heart disease, congestive heart failure, atrial fibrillation, asthma, chronic obstructive pulmonary disease, traumatic brain injury, vitamin B12 deficiency, depression, any anxiety disorder, nicotine dependence, and alcohol and drug abuse/dependence. We controlled for sustained use of medications that may impact cognitive function. Sustained use was defined as two prescriptions within any 6-month period in the 2 years prior to index. Medications included antidepressants, benzodiazepines, anticholinergics, NSAIDS, antihypertensives, statins, steroids, antivirals, metformin, and sulfonylurea.

Analytic approach

Receipt of vaccination is not random. Factors that are associated with vaccination, such as use of preventive health care, may also be linked to a lower dementia risk. To control for this and other sources of confounding, we computed propensity scores (PS) and inverse probability of treatment weighting (IPTW). Detailed PS and IPTW methods are reported in Appendix S1.

All primary analyses were performed with SAS v9.4 (SAS Institute, Cary, NC) at a two-tailed alpha = 0.05. Bivariate comparisons between covariates and vaccine group were assessed using chi-square tests and maximum ASMD% before and after IPTW. Cox proportional hazard models before and after weighting calculated hazard ratios and 95% confidence intervals for the association of vaccine group and incident dementia. Since mortality information was available in the VHA, competing-risk survival models were used to control for bias associated with a competing event (e.g., death).23 Weighted models used robust, sandwich-type variance estimators for confidence intervals and p values.24 The proportional hazard assumption was tested in all models with a time-dependent interaction term of vaccine group and log (follow-up time); the assumption was met for all models (p > 0.05).

To determine if unmeasured confounding could explain the current results, the e-values for the hazard ratio were calculated.25 The e-value is the minimum strength of association that is needed for an unmeasured confounder to have with both the exposure and outcome to completely explain observed associations. Since there were four exposure groups and multiple significant pairwise comparisons, the e-value for the largest significant point estimates was given for each cohort. This e-value would show the strength of association a confounder would need to explain all significant associations. We also employed a negative outcome control by modeling the association between vaccination and incident back pain in patients without back pain at index.26

Sensitivity analyses

To account for healthy adherer bias, we selected a subset of patients with hypertension at index (in VHA only, MarketScan sample size was insufficient) and expanded the final weighted model by adjusting for anti-hypertensive medication adherence, defined as ≥80% of proportion of days covered, in follow-up. Last, we adjusted for neighborhood socioeconomic index (nSES) measured at index date. Higher SES has been associated with a lower risk of dementia27 and is positively associated with vaccination.28-31 The nSES is a validated index32 created using 5-year census data from the American Community Survey and linked to zip codes (only available in VHA).

To determine if the association between vaccination and incident dementia was partly explained by tetanus, diptheria or pertussis, or HZ infection, we expanded weighted survival models to include two separate time-dependent variables for Tdap infection and HZ infection that could occur between index and end of follow-up.

RESULTS

As shown in Table 1, VHA patients were 76.8 ± 7.6 years of age, 4.4% were female, and 90.9% were White, most were married (69.5%), and 16.5% had access to only VHA health insurance. MarketScan patients were 70.5 ± 5.9 years of age. With exceptions for asthma and traumatic brain injury, comorbid conditions were more prevalent in VHA patients relative to MarketScan. VHA and MarketScan samples had similar prevalences of anticholinergic, NSAID, antidepressants, benzodiazepines, and antiviral use. Other medications measured were more prevalent among VHA patients.

TABLE 1. Baseline patient characteristics (%) of ≥65 years old patients in VHA (n = 80,070) and MarketScan (n = 129,200) Covariates VHA (n = 80,070) MarketScan (n = 129,200) Sociodemographic-related Age, mean (±SD) 76.8 (±7.6) 70.5 (±5.9) Age category 65–69 18,814 (23.5) 73,448 (56.9) 70–74 13,548 (16.9) 27,592 (21.4) ≥75 47,708 (59.6) 28,160 (21.8) Female gender 3540 (4.4) 84,480 (65.4) Race: White 72,761 (90.9) — Black 6404 (8.0) — Other 905 (1.1) — Married 55,673 (69.5) — VHA only insurance 13,216 (16.5) — Region Northeast 9353 (11.7) 51,329 (39.7) Northcentral 32,711 (40.8) 22,591 (17.5) South 28,846 (36.0) 32,879 (25.5) West 9160 (11.4) 22,401 (17.3) High healthcare utilizationa 20,004 (25.0) 32,351 (25.0) # well-visits 2 years prior to index, mean (±SD)b 1.0 (±2.1) 1.5 (±0.8) # well-visits 2 years prior to index, categoryb 0 45,262 (56.5) 14,186 (11.0) 1–2 26,124 (32.6) 104,921 (81.2) ≥3 8684 (10.9) 10,093 (7.8) Comorbidities Cancer 26,644 (33.3) 26,119 (20.2) Type II diabetes 26,594 (33.2) 20,424 (15.8) Obesity 26,070 (32.6) 7703 (6.0) Hypertension 66,803 (83.4) 78,550 (60.8) Stroke 3932 (4.9) 2350 (1.8) Ischemic heart disease 35,085 (43.8) 19,747 (15.3) Congestive heart failure 13,386 (16.7) 4426 (3.4) Atrial fibrillation 14,878 (18.6) 7847 (6.1) Asthma 6118 (7.6) 9697 (7.5) chronic obstructive pulmonary disease 18,094 (22.6) 8372 (6.5) Traumatic brain injury 2298 (2.9) 2316 (1.8) Vitamin B12 deficiency 4935 (6.2) 3600 (2.8) Depression 7143 (8.9) 4762 (3.7) Anxiety disorderc 6457 (8.1) 4153 (3.2) Nicotine dependence 20,488 (25.6) 6341 (4.9) Alcohol abuse/dependence 3173 (4.0) 570 (0.4) Drug abuse/dependence 895 (1.1) 198 (0.2) Medicationsd Anticholinergics 9325 (11.7) 14,532 (11.3) NSAIDS 9098 (11.4) 12,032 (9.3) Antihypertensives 56,107 (70.1) 64,578 (50.0) Statins 45,003 (56.2) 49,736 (38.5) Steroids 8261 (10.3) 7293 (5.6) Antivirals 903 (1.1) 2479 (1.9) Metformin 9757 (12.2) 9274 (7.2) Sulfonylurea 9412 (11.8) 4574 (3.5) Antidepressants 11,736 (14.7) 16,345 (12.7) Benzodiazepines 4348 (5.4) 7850 (6.1) Abbreviation: VHA, Veterans Health Administration.

Patient characteristics by vaccination status are shown in Table 2. Among VHA patients, 78.7% (n = 63,021) had no Tdap or HZ vaccination, 4.8% had Tdap only, 14.3% had HZ only, and 2.1% had both the Tdap and HZ vaccination by index. Among MarketScan patients, 78.8% had no Tdap or HZ vaccination, 7.6% had Tdap only, 10.7% had HZ only, and 2.9% had both the Tdap and HZ vaccination at index.

TABLE 2. Baseline patient characteristics (%) of ≥65 years old patients by vaccine status, VHA (n = 80,070) and MarketScan (n = 129,200) VHA patient cohort MarketScan patient cohort No vaccine n = 63,021 Tdap only n = 3874 HZ only n = 11,434 Tdap + HZ n = 1741 ASMD% No vaccine n = 101,819 Tdap only n = 9816 HZ only n = 13,774 Tdap + HZ n = 3791 ASMD% Covariates Age category 65–69 14,169 (22.5) 1571 (40.6) 2320 (20.3) 754 (43.3) 51.0 55,759 (54.8) 6785 (69.1) 8060 (58.5) 2844 (75.0) 43.4 70–74 10,328 (16.4) 611 (15.8) 2307 (20.2) 302 (17.4) 11.5 22,228 (21.8) 1711 (17.4) 3104 (22.5) 549 (14.5) 20.9 ≥75 38,524 (61.1) 1692 (43.7) 6807 (59.5) 685 (39.4) 44.6 23,832 (23.4) 1320 (13.5) 2610 (18.9) 398 (10.5) 34.9 Female gender 2711 (4.3) 207 (5.3) 514 (4.5) 108 (6.2) 8.5 67,395 (66.2) 5957 (60.7) 8816 (64.0) 2312 (61.0) 11.5 Race White 56,616 (89.8) 3442 (88.8) 11,061 (96.7) 1642 (94.3) 30.9 – – – – – Black 5681 (9.0) 385 (9.9) 267 (2.3) 71 (4.1) 32.1 – – – – – Other 724 (1.2) 47 (1.2) 106 (0.9) 28 (1.6) 6.1 – – – – – Married 42,977 (68.2) 2447 (63.2) 8998 (78.7) 1251 (71.9) 34.7 – – – – – VHA only insurance 11,020 (17.5) 862 (22.3) 1043 (9.1) 291 (16.7) 36.7 – – – – – Region Northeast 7260 (11.5) 512 (13.2) 1260 (11.0) 321 (18.4) 21.1 43,355 (42.6) 2868 (29.2) 4227 (30.7) 879 (23.2) 42.2 Northcentral 23,604 (37.5) 1432 (37.0) 6832 (59.8) 843 (48.4) 46.8 16,786 (16.5) 1938 (19.7) 3037 (22.1) 830 (21.9) 14.1 South 25,199 (40.0) 1285 (33.2) 2161 (18.9) 201 (11.6) 68.8 26,741 (26.3) 2092 (21.3) 3327 (24.2) 719 (19.0) 17.5 West 6958 (11.0) 645 (16.6) 1181 (10.3) 376 (21.6) 31.1 14,937 (14.7) 2918 (29.7) 3183 (23.1) 1363 (35.9) 50.5 High healthcare utilization 15,496 (24.6) 1553 (40.1) 2310 (20.2) 645 (37.1) 44.4 25,555 (25.1) 2395 (24.4) 3502 (25.4) 899 (23.7) 4.0 # well-visits 2 years prior to index 0 35,698 (56.6) 2318 (59.8) 6301 (55.1) 945 (54.3) 11.2 12,116 (11.9) 716 (7.3) 1125 (8.2) 229 (6.0) 20.6 1–2 20,380 (32.3) 1131 (29.2) 4075 (35.6) 538 (30.9) 13.8 82,197 (80.7) 8173 (83.3) 11,362 (82.5) 3189 (84.1) 8.9 ≥3 6943 (11.0) 425 (11.0) 1058 (9.3) 258 (14.8) 17.2 7506 (7.4) 927 (9.4) 1287 (9.3) 373 (9.8) 8.8 Comorbid conditions Cancer 21,374 (33.9) 1130 (29.2) 3696 (32.3) 444 (25.5) 18.5 20,458 (20.1) 1944 (19.8) 2954 (21.5) 763 (20.1) 4.1 Type II diabetes 21,333 (33.8) 1353 (34.9) 3416 (29.9) 492 (28.3) 14.4 16,599 (16.3) 1392 (14.2) 1954 (14.2) 479 (12.6) 10.4 Obesity 19,736 (31.3) 1552 (40.1) 4047 (35.4) 735 (42.2) 22.8 6059 (5.9) 628 (6.4) 753 (5.5) 263 (6.9) 6.1 Hypertension 52,786 (83.8) 3244 (83.7) 9396 (82.2) 1377 (79

留言 (0)

沒有登入
gif